High-fidelity modelling of a six-turbine tidal array in the Shetlands

Abstract

Micro-siting tidal stream turbines in a confined seabed area requires a extensive understanding of the flow dynamics over the water column at turbine deployment locations so that operating conditions are assessed, wake effects can be estimated to infer the energy yield [1], or bathymetry effects can be quantified. Tidal currents have the advantage of being highly predictable and mostly bidirectional but the uneven bathymetry found at most of sites introduces a high variability to the flow conditions within relatively short distances. Considering future tidal turbine arrays will comprise dozens of devices, deploying ADCPs at each turbine position would be very expensive or very time consuming, outlining the need for accurate modelling tools to be used as digital twins in micro-siting. Shallow-water models are widely adopted in preliminary design of tidal arrays but fail to capture the three dimensional nature of the flow and predict deflected wakes whose streamwise length is also over-predicted [2]. Thus, eddy-resolving method are required to fully capture the turbulence from the free-stream flow, induced by turbines and from bathymetry. \\

This study provides a real-project application of the state-of-the-art large-eddy simulation (LES) code DOFAS (Digital Offshore Farms Simulator) [3] to the six 100kW-turbine array deployed by NOVA Innovation Ltd. in the Shetlands, UK [4]. The bathymetry data has been obtained from EMODnet database with the velocity profiles set at the inlet condition of both ebb and flood tides imported from Macleod et al. (2019) [4]. The deployment site is characterised by steep slopes with a maximum depth of 45 m at the cross-section where turbines are located. The bathymetry has a downwards slope when the flow goes in the ebb tide direction whilst upstream during flood tide. The tidal rose indicated a slight deviation of about 20$^{\circ}$ between ebb and flood directions. The turbines have a diameter ($D$) of 9 m attached to a 10 m long hub whose diameter is 1 m. Three array configurations have been studied: (i) single row of three turbines, (ii) two rows of turbines spaced 8$D$, and (iii) two rows of turbines spaced 12$D$. The computational domain extends over 600 m by 300 m in the horizontal plane yielding 540 million cells, requiring 32,500 CPU hours to compute 30 min of physical time.\\

Results show that bathymetry effects at this site play a larger role when designing the location of the secondary row of turbines compared to wake effects from upstream turbines. During the ebb tide, the increase in water depth reduces the wake recovery far downstream so that the array with 12$D$ row spacing has a lower performance than the 8$D$ one with approx. 30\% decrease in energy yield. Conversely, the uphill shape of the bathymetry during the flood enables a fast wake recovery so that the downstream row experiences almost no energy yield loss due to wakes.